able to manage llm service access info(CRD LLM)

[bjwswang]

able to manage llm service access info(CRD LLM)

[Lanture1064]

Assumed Field:

metadata:
  name:
  namespace:
  label:
  annotation:
spec:
  modelName:
    description: 'modelName refers to the name of the model'
  modelVersion: v1
  modelSource:
    description: 'modelSource refers to the source of the llm, either "Local" as locally running model, or "Remote" as remote running model or commercial service'
  url:
    description: 'url refers to the url link of the llm'      
  auth:
    description: 'auth refers to the auth key to the llm, such as api key'
status:
  phase:
    description: 'phase marks the llm's running phase, either "running", "stopped", or "unavailable"'
  prompt:
    description: 'prompt is the last phrases sent to the llm, which should be an object from CRD Prompt'
  message:
    description: 'message is the returning result of last conversation'
  lastUpdateTime:
  retryCount:
parameters:
  description: 'parameters is the setting of the running model'
  maxTokens:
    description: 'maxTokens marks the limit of the input token quantity. 1 word averagely equals to 1.5 tokens.'
  temperature:
    description: 'temperature marks the 'randomness' of the model. Higher temperature, wilder models'
  otherParams:
    description: 'TODO: different model might have different params, such as fine-tuning params like top_K, plugin info, etc, which should be considered and dealt with'

[bjwswang]

@Lanture1064 Thanks! For the 1st version, we can only suppport openai protocols